Abstract:
|
We propose multiplicity adjusted evidence weights (MAEWs) as a general statistical method for evaluating the association of SNP genotypes with a broad variety of phenotypes. For each variant, the MAEWs indicate the strength of statistical evidence supporting each of four genetic models (null, additive, dominant, recessive) for the association of genotype with phenotype. MAEW addresses multiple-testing by adjusting Akaike Information Criterion (AIC) evidence weights for the empirical Bayesian probability (EBP) that the null hypothesis is true. Unlike p-values, MAEW provides a readily interpretable quantitative metric of the evidence supporting the four specific genetic models. We show that MAEW stringently controls the false discovery rate and asymptotically selects the correct genetic model for every variant. Also, in two-stage studies, using MAEW to select a specific model in the discovery stage improves the power and rate of correct validations in the validation stage. MAEW can also be used in settings where it is necessary to adjust for environmental factors. These properties are observed across many traditional simulation studies and in the GAW17 data set.
|